Contributions to Design and Analysis of Pediatric HIV Studies

Abstract

Youth with perinatal HIV exposure have demonstrated higher rates of emotional---behavioral problems than the general US youth population. However, more evidence is needed to help target prevention and intervention efforts throughout their lives. To address this need, we present three studies that provide novel insights on (1) racial/ethnic disparities in emotional---behavioral functioning, (2) identifying critical windows of exposure, and (3) sample size requirements for cluster randomized trials (CRTs) with binary baseline and follow-up measurements. Many of these findings are based on data from the Pediatric HIV/AIDS Cohort Study, one of the largest US-based cohort studies of youth with perinatal HIV exposure. In Chapter 1, we longitudinally examined racial/ethnic disparities in emotional---behavioral functioning among youth with perinatal HIV exposure. We applied mixed-effects models to evaluate whether mean youth-reported emotional concerns and caregiver-reported behavioral concerns differed by race/ethnicity. We also employed group-based trajectory models to identify groups having similar emotional---behavioral trajectories, followed by multinomial models to determine which factors predicted group membership. In Chapter 2, we present a statistical evaluation of multiple informant models implemented using generalized estimating equations (MIM GEEs). This approach has been used to identify critical windows of exposure. We evaluated the impact of correlation between exposure measurements and missing exposure data on the power and differences in association estimated by the MIM GEE and an inverse probability weighted extension to account for informatively missing exposures. We assessed these operating characteristics under a variety of correlation structures, sample sizes, and missing data mechanisms considering various exposure-outcome scenarios. We then applied these methods to a study of pregnant women living with HIV to explore differences in association between trimester-specific viral load and infant neurodevelopment. In Chapter 3, we developed sample size formula for CRTs with a binary post-test score conditioning on a binary pre-test score, and compare efficiency to CRTs that use only binary post-test scores. We propose a method to decrease the variance of the estimated treatment difference and thereby decrease the necessary sample size. Our work is developed within a semi-parametric framework that does not require the assumptions of maximum likelihood, while allowing for within-period, between-period, and within-individual correlations.